US 20070195285 A1 Abstract The present invention relates to projection systems where multiple projectors are utilized to create respective complementary portions of a projected image. More particularly, according to one embodiment of the present invention, a method of calibrating a multi-projector image display system is provided. According to the method, non-parametric calibration data for the display system is recovered and used to generate a non-parametric model of the display system. Local parametric models relating to the display surface of the projection screen are generated using canonical surface data representing the image projection screen. The local parametric models are compared with data points defined by the non-parametric calibration data to identify one or more local errors in the non-parametric calibration data. The local errors in the non-parametric calibration data are converted to data points defined at least in part by the local parametric models and the projectors are operated to project an image on the image projection screen by utilizing a hybrid calibration model comprising data points taken from the non-parametric model and data points taken from one or more local parametric models. Additional embodiments are disclosed and claimed.
Claims(22) 1. A method of calibrating an image display system comprising a plurality of projectors oriented in the direction of an image projection screen, the method comprising:
recovering non-parametric calibration data for the display system; generating a non-parametric model of the display system from the non-parametric calibration data; generating local parametric models relating to the display surface of the projection screen using canonical surface data representing, at least in part, the image projection screen; comparing the local parametric models with data points defined by the non-parametric calibration data; using the comparison of the local parametric models with the data points defined by the non-parametric calibration data to identify one or more local errors in the non-parametric calibration data; converting the local errors in the non-parametric calibration data to data points defined at least in part by the local parametric models; and operating the projectors to project an image on the image projection screen by utilizing a hybrid calibration model comprising data points taken from the non-parametric model and data points taken from one or more local parametric models to convert the local errors in the non-parametric calibration data. 2. A method as claimed in 3. A method as claimed in 4. A method as claimed in 5. A method as claimed in 6. A method as claimed in 7. A method as claimed in 8. A method as claimed in 9. A method as claimed in 10. A method as claimed in 11. A method as claimed in 12. A method as claimed in 13. A method as claimed in 14. A method as claimed in 15. A method as claimed in 16. A method as claimed in 17. A method as claimed in 18. A method of calibrating an image display system comprising a plurality of projectors oriented in the direction of an image projection screen, the method comprising:
utilizing one or more cameras oriented to take one or more images of the image projection screen to recover non-parametric calibration data for the display system from the geometry of the projection screen and the respective positions and orientations of the projectors; generating a non-parametric model of the display system from the non-parametric calibration data; generating local parametric models relating to the display surface of the projection screen using canonical surface data representing, at least in part, the image projection screen; comparing the local parametric models with data points defined by the non-parametric calibration data by registering each projector pixel corresponding to the non-parametric data points to a canonical surface approximating the geometry of the image projection screen and determining the distance between a match point of the projector pixel and the local model. using the comparison of the local parametric models with the data points defined by the non-parametric calibration data to identify one or more local errors in the non-parametric calibration data as a function of the distance between the match point and the local model; converting the local errors in the non-parametric calibration data to data points defined at least in part by the local parametric models by replacing perturbed local data points within the non-parametric model with a corresponding point generated by one of the local parametric models when the comparison with the local parametric models is indicative of a surface abnormality in the image projection screen, error in an estimated camera position, or differences in the canonical model and the true projection screen geometry such that the method of calibration is fundamentally a non-parametric system that incorporates parametric constraints in local regions to correct user-detectable artifacts or discontinuities in the non-parametric calibration model and such that global problems typically associated with parametric calibration data are avoided, while retaining the local consistency that parametric models provide; operating the projectors to project an image on the image projection screen by utilizing a hybrid calibration model comprising data points taken from the non-parametric model and data points taken from one or more local parametric models to convert the local errors in the non-parametric calibration data; and verifying the hybrid calibration model by generating an image of a 3D calibration mesh configured such that errors in the hybrid calibration model can be identified by a user or one or more image analysis cameras when the calibration mesh image is displayed on the projection screen. 19. A method of operating a display system in accordance with an image rendering algorithm to project an image on an image projection screen utilizing a plurality of projectors oriented in the direction of the image projection screen, the image rendering algorithm incorporating a hybrid calibration model generated by methodology comprising:
recovering non-parametric calibration data for the display system; generating a non-parametric model of the display system from the non-parametric calibration data; generating local parametric models relating to the display surface of the projection screen using canonical surface data representing, at least in part, the image projection screen; comparing the local parametric models with data points defined by the non-parametric calibration data; using the comparison of the local parametric models with the data points defined by the non-parametric calibration data to identify one or more local errors in the non-parametric calibration data; converting the local errors in the non-parametric calibration data to data points defined at least in part by the local parametric models; and establishing the hybrid calibration model such that it comprises data points taken from the non-parametric model and data points taken from one or more local parametric models to convert the local errors in the non-parametric calibration data. 20. A method as claimed in 21. A method of calibrating an image display system comprising a plurality of projectors oriented in the direction of an image projection screen and at least one calibration camera, wherein at least the following conditions apply to the method:
the calibration camera captures k distinct images of the image projection screen; all projectors contributing to each captured image render a set of fiducials captured by the calibration camera; a set of three-dimensional points corresponding to camera image points are computed as respective intersections of back-projected rays defined by the points and a canonical surface approximating the projection screen; the points are matched with projected fiducials to generate a set of corresponding match points; the set of three-dimensional points observed in different camera views are represented as a set of 3D surface points with a known neighborhood function; the 3D points are modeled as a constraint system such that the error distance between two points seen in two different camera views are computed as the geodesic distance between the first point, as seen in the second view, and the second point, as seen in that same view; and points that correspond to the same projector location but have different locations on the 3D surface are adjusted according to an error metric that minimizes the total error represented in the constraint system. 22. A method as claimed in Description This application claims the benefit of U.S. Provisional Application Ser. No. 60/773,419, filed Feb. 15, 2006. The present invention relates to projection systems where multiple projectors are utilized to create respective complementary portions of a projected image, which may be a video or still image. More particularly, the present invention relates to methods of calibrating and operating such systems. According to one embodiment of the present invention, a method of calibrating a multi-projector image display system is provided. According to the method, non-parametric calibration data for the display system is recovered and used to generate a non-parametric mapping of positions in each projector to their position within a common global reference frame of the display system. Local parametric models that relate to the display surface are generated using a canonical description that either represents the image projection screen or the expected position of neighboring points when projected onto the screen. In addition, these local parametric models may represent the expected position of points in one device, e.g., a projector, when they are known in a second device, e.g., a camera. These local parametric models are compared with data points defined by the non-parametric calibration data to identify one or more local errors in the non-parametric calibration data. The local errors in the non-parametric calibration data are converted to data points by referring, at least in part, to the local parametric models. Although the conversion may be solely a function of the parametric model, it is contemplated that the conversion may be a function of both the parametric model and the non-parametric mapping, e.g., by referring to the predicted data points given by the parametric models and measurements taken from the non-parametric mapping. The projectors are operated to project an image on the image projection screen by utilizing a hybrid calibration model comprising data points taken from the non-parametric model and data points taken from one or more local parametric models. In accordance with another embodiment of the present invention, a method of operating a multi-projector display system is provided. According to the method, the display system is operated according to an image rendering algorithm that incorporates a hybrid parametric/non-parametric calibration model. In accordance with another embodiment of the present invention, a method of calibrating an image display system is provided. The system comprises a plurality of projectors oriented in the direction of an image projection screen and at least one calibration camera. According to the method, the calibration camera captures k distinct images of the image projection screen. All projectors contributing to each captured image render a set of fiducials captured by the calibration camera. A set of three-dimensional points corresponding to camera image points are computed as respective intersections of back-projected rays defined by the points and a canonical surface approximating the projection screen. The points are matched with projected fiducials to generate a set of corresponding match points. The set of three-dimensional points observed in different camera views are represented as a set of 3D surface points with a known neighborhood function. The 3D points are modeled as a constraint system such that the error distance between two points seen in two different camera views are computed as the geodesic distance between the first point, as seen in the second view, and the second point, as seen in that same view. Points that correspond to the same projector location but have different locations on the 3D surface are adjusted according to an error metric that minimizes the total error represented in the constraint system. The following detailed description of specific embodiments of the present invention can be best understood when read in conjunction with the following drawings, where like structure is indicated with like reference numerals and in which: Generally, various embodiments of the present invention relate to calibration techniques that utilize local parametric models in conjunction with global non-parametric models. Although the calibration methodology of the present invention has broad applicability to any image projection system where an image or series of images are projected onto a viewing screen, the methodology of the various embodiment of the present invention are described herein in the context of a complex-surface multi-projector display system referred to as the Digital Object Media Environment (DOME). Referring to Referring to Referring now to the flow chart of Once the non-parametric model has been established (see step The calibration scheme illustrated in Once this global, non-parametric model has been acquired (see step Thus, in the multi-projector calibration scheme illustrated in For example, and not by way of limitation, in the multi-projector system Given the assumption that observed points in the camera plane arise from projected fiducials on the canonical surface, then the three-dimensional point [x y z] Because the canonical surface in the case of the DOME The nine parameters of P can be recovered via a robust least squares fit, for a given match point over a 5×5 grid of neighboring points. Typically, the match point under consideration is not used during the fit. Instead, the distance between the match point and the fit model P is measured and if this distance exceeds some threshold, the match point is considered to be in error, and is discarded. The local parametric model is then used to interpolate a new match point at this location. This set of three-dimensional points observed in different camera views must be registered to a single three-dimensional point cloud. If the same projector point is seen in multiple views only one is selected by iterating through multiple camera views and adding only unique points until the point cloud is fully populated. Next, a 3D Delaunay triangulation is performed on this point cloud to compute neighbor relations. Finally, this 3D mesh is modeled as a constraint system in which each edge is assigned a weight of one and a length, i.e., an error distance, that corresponds to the separation of the two points on the sphere. In the case when two points arise from the same camera view, the distance is equivalent to the geodesic distance. However, if the two points p Following error distance assignments, the constraint model is relaxed in order to minimize the total error contained in the constraint system. This minimization phase may use a variety of minimization techniques including traditional gradient, downhill simplex, simulated annealing, or any other conventional or yet to be developed energy minimization technique. As a result, local errors are distributed over the mesh, including those arising from error propagation between views, error in estimated camera positions, improperly modeled radial distortion, etc. This yields a perceptually consistent calibration across all projectors Once the projectors At each frame, the head-positions of the viewers The rendered view for each projector It is noted that recitations herein of a component of the present invention being “configured” to embody a particular property, function in a particular manner, etc., are structural recitations, as opposed to recitations of intended use. More specifically, the references herein to the manner in which a component is “configured” denotes an existing physical condition of the component and, as such, is to be taken as a definite recitation of the structural characteristics of the component. It is noted that terms like “preferably,” “commonly,” and “typically” are not utilized herein to limit the scope of the claimed invention or to imply that certain features are critical, essential, or even important to the structure or function of the claimed invention. Rather, these terms are merely intended to highlight alternative or additional features that may or may not be utilized in a particular embodiment of the present invention. Having described the invention in detail and by reference to specific embodiments thereof, it will be apparent that modifications and variations are possible without departing from the scope of the invention defined in the appended claims. More specifically, although some aspects of the present invention are identified herein as preferred or particularly advantageous, it is contemplated that the present invention is not necessarily limited to these preferred aspects of the invention. For example, although the calibration methodology of the present invention has been described herein in the context of a complex-surface multi-projector display system referred to as the Digital Object Media Environment (DOME), the appended claims should not be limited to use with the DOME or similar projection systems unless they expressly recite the DOME. Referenced by
Classifications
Legal Events
Rotate |